September 27, 2011
The assault on hard disk storage seems to be building with each new flash memory offering. This week, Violin Memory launched a new solid state memory line aimed to replace primary storage in the datacenter. The Violin 6000 Series flash Memory Arrays is designed as an all-silicon storage solution for data-intensive enterprise applications, and is intended to compete against disk-based solutions in cost, both upfront and operationally.
The company also added to its 3200 Series flash lineup. Geared for maximum performance, capacity, and flexibility, the 3000 offerings are slightly lower on the storage food chain than the new 6000 line. The new entry, the 3U Violin 3220, doubles the storage capacity of the 3200 Series from 10 TB to 20 TB (16 TB usable) and delivers 250,000 IOPS.
The 6000 platform is the bigger news for the company though. It incorporates high-end enterprise features such as reliability/redundancy, high availability, serviceability, and scale-out management. As such, it extends Violins storage reach into high-end use cases for tier 1 storage, opening up a potentially much larger market. The 3U product comes in two flavors: the 6216 performance array (1M IOPS; 16 TB SLC flash, 12 TB usable) and the 6632 capacity array (500K IOPS; 32 TB MLC flash, 22 TB usable). For write endurance, the SLC-based 6216 would be the logical choice at 800 PB of writes, versus the 6632 at 100 PB.
Violin's new 6000 and 3220 highlights the company's focused strategy to displace traditional disk-based storage in the enterprise. Unlike many SSD vendors, who are primarily offering a high performance tier 0 cache between disk storage and memory, Violin is looking to penetrate more deeply into the datacenter. "Across our product line we believe we can deliver both lower CAPEX and OPEX for tier 0, tier 1 and tier 2 applications," says Violin CEO Don Basile "We think we can remove those disks from the datacenter and end up with what we call a flash and trash strategy."
A number of industry analysts, not to mention storage vendors, are touting the so-called "flash and trash" strategy, a two-tiered storage model, in which solid state drives are used for all active storage and cheap, high capacity SATA disks for data that needs saving, but not quick access. Basile argues that such a model has two big benefits: First, replacing spinning disk with solid state flash saves a boatload of money by reducing floor space and power usage; and second, storage infrastructure is simplified enormously by flattening all the active data tiers into a single layer.
Arun Taneja, founder and analyst at Taneja Group, thinks Violin is paving the path to eliminating HDD-based systems but it will take three or more years for that to happen. He also believes at least some customers are going to need other high-end features, like replication and snapshot capabilities, integrated into the storage system, which is something the initial 6000 boxes don't support yet. "Rome wasn't built in a day," notes Taneja. "The fact that Violin is involved in building Rome is what is really important, in my view."
Meanwhile, Violin is picking off a lot of low-hanging fruit, especially where I/O performance is paramount. The company was in a recent bake-off in which the customer required 40 TB (usable) with 500,000 I/Os. According to Basile, for traditional vendors that mean six racks of disk, with sub-20 percent utilization. The upfront cost for the disk-based solution was about $3 million, plus more than $3 million more for operational expenses. The customer ended up with the Violin solution, a 3000 series flash array. It delivered the requisite 40 TB, along with much more than 500K I/Os at sub-$1 million CAPEX and sub-$1 million OPEX.
"That's what our global CIOs are looking at as they make this transformation from these giant beasts that ruled the earth -- these dinosaurs -- to a more nimble viable infrastructure that scales," says Basile. " This is where we think the memory array is not a slightly better mousetrap, but a way to catch mice differently."
Violin certainly seem to be on the right track. Since the company launched its first solid state storage product into the market just 13 months ago, they've been in growth-only mode. For the fiscal year ending in January, they're expecting to take in $100 million in revenue. According to Basile, that trajectory places them about 4 years ahead of Isilon or 3PAR during their initial ramp up. "Our growth rates are faster than any storage company in history on a revenue basis," says Basile, who as the former CEO of Fusion-io, must have a pretty good idea of what fast starts look like.
As a privately held company, Violin is living off $110 million in funding from a number of investors, the biggest ones being Toshiba (the maker of Violin's flash memory components), Juniper Networks, and an unnamed system/storage vendor. They might be getting close to cashing in though. The company's growth spurt has them thinking about a public offering, and, according to Basile, they are looking to file an IPO in Q1 or Q2 of 2012.
The Violin CEO thinks 2012 is shaping up to be a transition year, in which a lot of big enterprise accounts will move their data from disk to memory arrays as a core storage tier. The company is trying to quickly gear up to catch that wave, piggybacking on partner companies like IBM and HP to help reach those customers and manufacturing partner Flextronics to churn out product.
With a current head count of just over 220 employees, Violin expects to at least double that by the first quarter of 2012.
Violin's three biggest industry targets are financial services, government, and internet properties. In partnership with HP, they've produced an alternative to Oracle's popular Exadata database (DB) appliance. Besides vanilla database and analytics work, Violin storage can be used for a range of "big data" applications including backtesting trading strategies for Wall Street firms, accelerating advertising analytics for companies like AOL (where a petabyte of Violin storage is deployed), and serving up real-time advertisements to mobile clients.
In the government space, primarily US defense and intelligence, the use cases are not for public consumption, but one can safely assume that the Violin gear is behind some serious, high-IOPS data mining. The largest deployment there is just over a petabyte for an unnamed US government agency.
For HPC-type applications, the company is increasingly looking at Hadoop clustering and Cassandra/BigTable applications. In addition, they teamed with IBM to demonstrate record-breaking file system performance of Big Blue's General Parallel File System (GPFS), using Violin 3205 flash arrays as the metadata accelerator. HPC in general seems a rich opportunity for Violin storage, especially for their 3000 series memory boxes.
The new 3220, 6216, and 6632 products are already shipping, with the new 3200 box generally available. In the case of the 6000 offerings, only Violin's strategic partners and customers have access to them for the time being, with general availability slated for Q1 2012. Basile says they can't make enough of them right now to keep up with demand, adding, "we've pretty much sold all of them for the next 12 weeks."
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?